|
Activity Number:
|
287
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #310129 |
|
Title:
|
Bayesian Model Selection in High-Dimensional Genetic Association Studies
|
|
Author(s):
|
Melanie Wilson*+
|
|
Companies:
|
Duke University
|
|
Address:
|
Box 90251, Durham , NC, 27708,
|
|
Keywords:
|
Genetic Association Studies ; Bayesian Model Selection ; Hierarchical Models
|
|
Abstract:
|
Modern genotyping techniques allow vast amounts of data to be collected for genetic association studies. With this increase in data comes an increased need for statistical models that are able to sort through a large number of correlated covariates as possible disease predictors. To date, the results of such genetic association studies have been disappointing due to the focus on marginal association that over-simplifies the complex etiology of common disease. It is clear that studies of genetic variation and disease must account for the structure and function of genetic pathways. We propose Bayesian hierarchical model selection techniques that search gene-environment and gene-gene interactions in a computationally efficient manor while aiming to strike a balance between model complexity and analytical simplicity. We then compare methods using simulated and real-world datasets.
|
